System Architecture

The Buddi system architecture is designed as an integrated, modular, and scalable ecosystem that seamlessly combines affective computing, ambient sensing, and adaptive interaction design. Its architecture ensures real-time emotional support, personalized engagement, and privacy-preserving classroom analytics.

1. Overview

At a high level, Buddi consists of four interconnected layers:

  1. User Interaction Layer – Interfaces through which students engage with Buddi.
  2. Companion Intelligence Layer – Core logic managing the virtual companion, emotional modeling, and adaptive behavior.
  3. Intervention & Engagement Layer – Micro-interventions, gamification modules, and real-time feedback mechanisms.
  4. Data Aggregation & Analytics Layer – Classroom-level insights, emotion trend mapping, and secure data management.

Each layer communicates through API endpoints and event-driven pipelines, ensuring low-latency response while maintaining data privacy and modularity.

2. Layer Details

A. User Interaction Layer

  • Components:
    • Virtual Companion UI: A playful, responsive creature that mirrors the student’s mood, engagement, and habits.
    • Micro-Intervention Widgets: Sensory-friendly exercises, guided breathing, doodling, and grounding gestures.
    • Gamified Focus Trails: Progress bars, constellations, and soft visual cues encouraging task completion and emotional regulation.
  • Technology Stack:
    • Frontend: React Native / Flutter for cross-platform support.
    • UI/UX Framework: Material Design + soft gamification elements.
  • Responsibilities:
    • Capture student interactions (touch, click, duration, engagement).
    • Provide immediate feedback based on the Companion Intelligence Layer.
    • Ensure low-stimulation, accessible design for neurodivergent learners.

B. Companion Intelligence Layer

  • Components:
    • Emotion Recognition Engine: Uses ambient sensor data (optional wearables, keyboard/mouse patterns, facial expressions) to infer emotional states.
    • Behavior Modeling Module: Maintains a dynamic profile of each student’s habits, mood patterns, and responsiveness.
    • Adaptive Response Generator: Decides appropriate companion reactions, micro-interventions, and encouragement prompts.
  • Technology Stack:
    • Backend: Python / Node.js
    • ML Models: Lightweight affective computing models (TensorFlow Lite or PyTorch Mobile)
  • Responsibilities:
    • Translate raw sensor input into quantitative emotional metrics.
    • Update companion behavior dynamically to ensure meaningful, empathetic interactions.
    • Trigger context-aware micro-interventions without disrupting classroom flow.

C. Intervention & Engagement Layer

  • Components:
    • Micro-Intervention Scheduler: Determines timing and type of interventions based on emotional and engagement data.
    • Gamification Engine: Generates soft rewards, progress visualizations, and focus trails to motivate self-regulation.
    • Feedback Loop: Collects interaction data to refine intervention effectiveness using reinforcement learning.
  • Technology Stack:
    • Event-driven backend using WebSockets / MQTT for real-time updates.
    • Gamification logic implemented in frontend with server validation for consistency.
  • Responsibilities:
    • Encourage habit formation and consistent emotional regulation.
    • Provide instant, playful feedback to sustain engagement.
    • Avoid overstimulation or intrusive notifications.

D. Data Aggregation & Analytics Layer

  • Components:
    • Classroom Mood Dashboard (EchoGarden): Visualizes aggregated mood patterns across the classroom without exposing individual identities.
    • Analytics & Reporting Engine: Generates trend analysis, engagement reports, and intervention effectiveness metrics for educators.
    • Secure Storage & Privacy Layer: Ensures data encryption, anonymization, and compliance with privacy standards.
  • Technology Stack:
    • Database: PostgreSQL / Firebase Realtime Database
    • Analytics: Python Pandas / Plotly / D3.js
  • Responsibilities:
    • Provide actionable insights to teachers while maintaining student privacy.
    • Identify stress hotspots or patterns of emotional overload.
    • Store longitudinal behavioral data to inform companion personalization over time.

3. Data Flow

  1. Input: Student interacts via UI or ambient sensors detect emotional cues.
  2. Processing: Companion Intelligence Layer analyzes inputs and determines emotional state.
  3. Intervention: Engagement Layer triggers micro-interventions or gamified feedback.
  4. Aggregation: Data is anonymized and sent to Analytics Layer for classroom-level insights.
  5. Feedback Loop: Insights refine behavior models, improving future companion responses and interventions.

Alt text

4. Security & Privacy Considerations

  • Data Anonymization: No personally identifiable information stored in aggregated analytics.
  • Encrypted Communication: All frontend-backend interactions are over TLS/HTTPS.
  • Local-First Processing: Emotion recognition and companion behavior logic can run locally on the device, minimizing cloud dependency.

5. Scalability

  • Modular design allows for:
    • Adding new interventions or gamification mechanics.
    • Extending to multiple classrooms or schools without significant backend reconfiguration.
    • Integration with wearables or IoT classroom sensors in the future.

In summary, Buddi’s architecture combines playful, adaptive interaction design with robust emotion analytics to create a scalable, privacy-conscious digital ecosystem that proactively supports neurodivergent learners while empowering educators with actionable, anonymized insights.

Next : Layers